2016 Postmortem
Related: About this forumTPM's poll average is garbage
They use 4 Rasmussen polls in their average. This is why Romney is ahead in their average, which is ludicrous. Whatever pro-Republican effect of Rasmussen is magnified. This idiotic methodology is not used by Pollster.com or Realclearpolitics.
http://polltracker.talkingpointsmemo.com/contests/us-president-12/
Drunken Irishman
(34,857 posts)longship
(40,416 posts)In fact, TPM's electoral map is one of the more pro-Obama projections. The electoral map that excludes Rasmussen polls has the same electoral votes for Romney and is the only electoral map that I've seen more positive for Obama.
I do not like throwing data out. No scientist would do that, biased or not. There are always biases. Throwing out data is never acceptable. That's why there are error bars and why statistics tells us how to calculate them.
Another issue here. If you disagree about methodology, you cannot merely criticize the outcome. You have to criticize the methodology itself. These polls publish their methods. Just because you do not like the results doesn't mean the poll's methods are biased.
Polling is a messy art. Probably, the only way to get accuracy is to take the data as a whole. If you arbitrarily throw our a Republican leaning outlier, you should probably throw out Democratic leaning polls, too. Of course, this would be a slippery slope. Where do you draw the line?
No. Do not throw out any data. Use it all. The biases will average out, at the cost of uncertainty. That's how the statistics work.
TPM's electoral map is fine, especially since they don't throw out data.
Drunken Irishman
(34,857 posts)It's trash. Rasmussen does a daily tracking poll, as does Gallup, and TPM includes outdated polling that is already irrelevant the second Rasmussen releases its newest polls.
It's not valid to keep a Rasmussen poll in the sample from July when they've already released other polls that show a far different outcome. It's not credible.
So, no, no one is saying throw out Rasmussen - they're saying throw out the outdated polls. RCP does this and Obama's average lead is +3.0 - far different from being down -.6. I think it's clear TPM is using a junk outline for its data collecting and it's flawed because Rasmussen's polls change daily.
CreekDog
(46,192 posts)Rasmussen's daily tracking is 1 new day of polling combined with 2 days of recycled sample.
so when you include it as if it were 3 unique samples, when it's not...it's junk. and misleading junk at that.
Marzupialis
(398 posts)I am not talking about a map. I am talking about the poll tracker. That's why I linked to it.
longship
(40,416 posts)I am using an iPhone, so maybe that's an issue. But I have not had such problems at TPM before.
on edit: when I couldn't click through I assumed you meant TPM's electoral map. My bad; your good.
Marzupialis
(398 posts)Because the link works on my PC right now.
CreekDog
(46,192 posts)It's not just that they count Rasumussen and Gallup multiple times, it's that they double and triple count the SAME poll respondents by doing this!
For example:
if you include:
Rasmussen 8/2 poll (3-day tracking)
Rasmussen 8/1 poll (3-day tracking)
Rasmussen 7/31 poll (3-day tracking)
as independent polls, you've included some poll respondents 3 TIMES, because it's a 3-day rolling average.
same with Gallup.
it's a huge problem.
Garbage. I would tell them at @TPM. And often.